Random-Projection-Filters/config.yml at main · UniSerj/Random-Projection-Filters · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
TRAIN:
    # Number of training epochs
    epochs: 120

    # Architecture name, see pytorch models package for
    # a list of possible architectures
    arch: 'resnet50'

    # Starting epoch
    start_epoch: 0

    # SGD paramters
    lr: 0.04
    momentum: 0.9
    weight_decay: 0.0001

    # Print frequency, is used for both training and testing
    print_freq: 500

    # Dataset mean and std used for data normalization
    mean: !!python/tuple [0.485, 0.456, 0.406]
    std: !!python/tuple [0.229, 0.224, 0.225]

ADV:
    # FGSM parameters during training
    clip_eps: 4.0
    fgsm_step: 4.0

    # Number of repeats for free adversarial training
    n_repeats: 4

    # PGD attack parameters used during validation
    # the same clip_eps as above is used for PGD
    pgd_attack:
    - !!python/tuple [10, 0.00392156862] #[10 iters, 1.0/255.0]
    - !!python/tuple [50, 0.00392156862] #[50 iters, 1.0/255.0]

    # PGD training setings
    delta_init: 'zero' # ['zero', 'random']

    attack_iters: 2

DATA:
    # Number of data workers
    workers: 4

    # Training batch size
    batch_size: 1024

    # Image Size
    img_size: 256

    # Crop Size for data augmentation
    crop_size: 224

    # Color value range
    max_color_value: 255.0